Claude Fable 5 moves to usage credits: Pro now pays per token
As of 19 July 2026, Fable 5 is no longer included in Pro and Team standard plans. A close read: how credits work, the $100 terms and the expiry dates.
On 19 July 2026 the promotional period that let subscribers use Claude Fable 5 within their plan limits ended. After that date the policy splits in two: Max and Team premium seats keep Fable 5 as a standard part of the plan; Pro and Team standard seats do not, so using it now means paying per token with usage credits, or upgrading to Max.
To soften the change, Anthropic issued a one-time $100 usage credit to the group that lost access. Simple on the surface, but the terms contain enough easy-to-miss detail to ruin a subscription decision. This is a policy read-through aimed at decisions, not a news recap.
Every date below is in PT and falls between July and September 2026. Anthropic plan policy moves quickly, so verify against the Claude Help Center before making a financial call.
Who still gets Fable 5 in-plan
The line is drawn by plan tier, not account type:
- ◆Max, Team premium seats, and premium seats on legacy seat-based Enterprise: Fable 5 is included, up to roughly 50% of the weekly limit. No credits needed, and not eligible for the $100.
- ◆Pro and Team standard seats: Fable 5 is not included. Either burn credits or move to Max. This is the only group eligible for the $100 credit.
- ◆Free: no Fable 5 at all.
Usage credits are not extra quota
This is the most misread part. Usage credits are an "extra usage" mechanism: once you exhaust your plan limit, instead of being blocked you keep going at standard API rates, billed against credits. Credits apply across every surface (claude.ai, Claude Code CLI or IDE, Claude Desktop) and all of that usage rolls into the same limit.
But when credits get drawn down differs by model, and that is the decisive detail:
- ◆Fable 5 on Pro/Team standard: credits are charged from the first token, because the model sits outside the plan and has no quota to spend first.
- ◆Opus / Sonnet / Haiku: credits are charged only after the plan limit is used up. While quota remains, credits sit untouched.
"Credits work with any model" is true as written, but they are an overflow buffer, not a parallel quota. If you rarely hit your Pro ceiling, the only realistic way to spend the $100 is on Fable 5.
A few operational details: plan limits still reset every 5 hours as usual, and spending credits does not alter that cadence; the Usage dashboard reports in-plan usage and credit consumption separately.
To spend credits at all you must add a payment method and enable usage credits on the web app. Accounts subscribed through a mobile app cannot enable it in-app, so it has to be done on the web.
The $100 terms: eligibility closed before the claim window opened
The complete terms:
- ◆Eligibility: an active Pro or Team standard subscription as of 19 July 2026, 23:59 PT. Free trials do not count.
- ◆Amount: Pro $100. Team: $100 per purchased standard seat, capped at $2,500 per organization.
- ◆Claim window: 20 July to 2 August 2026, 23:59 PT, under Settings → Usage → "Claim free credits".
- ◆Expiry: 17 September 2026, 23:59 PT, fixed, regardless of when you claim. Claiming late buys no extra days.
- ◆Scope: standard usage credits, valid on any model, not restricted to Fable 5.
- ◆Plan binding: the balance still displays after you cancel or downgrade, but it can only be spent while you remain on an eligible plan. Non-refundable, no rollover, one time only.
Why "buy one month of Pro and bank the credit" fails
The plan breaks at two independent layers; solving one leaves the other standing.
One, you are not eligible. Eligibility was measured on 19 July 2026, before the claim window opened. The 20 July to 2 August period is only the time to press the button for people who already qualified, not a window to subscribe into.
Two, even if you had it, you could not spend it. Credits are bound to remaining on an eligible plan. Let the subscription lapse and the balance stays on screen but stops being usable.
Spending the full $100 means keeping Pro continuously through 17 September 2026, roughly two more renewal cycles from the end of July. In other words, the "free" credit is a retention incentive, not a standalone bargain.
For Claude Code users
- ◆CLI version 2.1.170 or newer is required before Fable 5 appears as an option.
- ◆Usage credits apply to Claude Code exactly as they do to claude.ai; usage across every surface accumulates against the same limit.
- ◆A secondary promotion is running: Claude Code weekly limits are 50% above standard through 19 August 2026. It covers Pro, Max, Team, and legacy seat-based Enterprise, applies automatically, and needs no setup.
Pricing and the math
Fable 5 bills against credits at $10 per 1M input tokens and $50 per 1M output tokens. For reference, first-party API pricing at the same time: Opus 5 at $5/$25 per MTok, Sonnet 5 at $3/$15, making Fable 5 twice the price of the Opus tier.
$100 in credits translates to roughly 2M output tokens if you run nothing but Fable 5 and ignore input entirely. Real-world numbers land meaningfully lower, since every session consumes input, and for a coding agent working over a large codebase input is rarely a rounding error.
Two routes, chosen by how often you actually reach for it
- ◆Stay on Pro and top up usage credits: pay per use, keep costs visible. Sensible when Fable 5 is reserved for hard tasks rather than every session.
- ◆Move to Max: Fable 5 is included up to ~50% of the weekly limit, with no credit balance to think about. Sensible for steady, heavy use. The trade-off: Max holders were not eligible for the $100.
Quick heuristic: estimate the output tokens you genuinely push through Fable 5 per month. Below roughly 1M tokens a month, Pro plus credits is cheaper; consistently above that, Max almost always wins, and you buy the freedom of not watching a balance.
Sources: Claude Fable 5 on your plan, Claude Fable 5 one-time free credits promotion and Manage usage credits for paid Claude plans on the Claude Help Center.
AkiTao is documenting this change because it lands directly on the operating cost of the AI-native workflows we use daily to build and maintain content systems. Model selection is no longer a purely technical question; it is a cost line with an expiry date on it.